[llvm] [AArch64] Neoverse V2 FeatureDisableLatencySchedHeuristic (PR #140897)

[Sjoerd Meijer via llvm-commits]

sjoerdmeijer wrote:

Hi @igogo-x86 , sorry for the late reply, thanks for sharing this.
I have also benchmarked this patch on the V2. I don't think I saw this regression in x264, not sure if I missed this or if something else is going on. I trust your numbers, but I am going to double check them at my end just out of curiousity. I will report back if I find something interesting.  

We have certainly seen some up and down behaviour too. But there were a lot more ups than downs (and a big up), making this overall a good change. 

We do have a follow up task on our backlog, we are interested to see if we can claw back one of the regressions. But like you say, this is tricky business because of scheduler and regalloc interaction, and there are different ways of solving this, so we will have to see. I don't think I can get more concrete than this at this point. The x264 problem sounds a bit different though, so I am going to have a look at https://github.com/llvm/llvm-project/pull/136329 and see if I can help with that one.





https://github.com/llvm/llvm-project/pull/140897